Preprocessed Kubric Movi-F basically following CoTracker with additional depth and camera parameters.
Each scene contains the following structure when extracted:
0000/
├── depths
| ├── 000.png
| ├── ...
| └── 023.png
├── frames
| ├── 000.png
| ├── ...
| └── 023.png
└── 0000.npy
frames: RGB images in .png format
depths: Depth data in .png format and unint16 data type
annotation: annotation in .npy format… See the full description on the dataset page:
https://huggingface.co/datasets/ZhengGuangze/LBM_Kubric_v2.
